Let’s see how this new dataset looks:
conn <- dbConnect(RSQLite::SQLite(), "../../Results/yeast.sqlite")
replicationTiming <- dbGetQuery(conn, "SELECT l.Gene, x, y, z, ReplicationTiming FROM Loci l JOIN ReplicationTiming r ON l.Gene = r.Gene")
plot_ly(x=replicationTiming$x, y=replicationTiming$y, z=replicationTiming$z, type="scatter3d", mode="markers", text=replicationTiming$Gene, color=replicationTiming$ReplicationTiming)dbDisconnect(conn)We use standard deviation of replication timing in a sphere as a metric.
This signal is so smooth that almost all possible spheres have a far below metric than the randomized set of same size.
Maybe it makes more sense to find the rare spheres where smoothness is disrupted:
sql <- "SELECT l.*, CASE WHEN Field IS NULL THEN 'None' ELSE Field END AS Field FROM Loci l LEFT JOIN ReplicationTimingFields f ON l.Gene = f.Gene"
fieldColors <- c("#FF0000", "#00FF00", "#0000FF", "#FFFF00", "#00FFFF", "#4488AA", "#FF00FF", "#000000")
conn <- dbConnect(RSQLite::SQLite(), "../../Results/yeast.sqlite")
replicationTimingFields <- dbGetQuery(conn, sql)
plot_ly(x=replicationTimingFields$x, y=replicationTimingFields$y, z=replicationTimingFields$z, type="scatter3d", mode="markers", text=replicationTimingFields$Gene, color=replicationTimingFields$Field, colors=fieldColors)dbDisconnect(conn)Field C is the same as MotifFields A, which was again the weird one, having more variance in TFs than expected by chance. These 123 genes are all located in chrXII, from ~300Kb to ~800KB, with a 200Kb gap between ~500-700.
Maybe we have the wrong (index space) coordinates for these genes? That would explain having significantly varied signals for both signal types.
sql <- "SELECT Gene, Chromosome, Start,End FROM Loci WHERE Gene IN (SELECT Gene FROM ReplicationTimingFields WHERE Field = 'C') ORDER BY Start"
conn <- dbConnect(RSQLite::SQLite(), "../../Results/yeast.sqlite")
weirdGenes <- dbGetQuery(conn, sql)
dbDisconnect(conn)
weirdGenes